<ul><li>Neural operators are used for learning solution operators of partial differential equations.</li><li>To address issues in time-dependent problems, Recurrent Neural Operators (RNOs) framework has been proposed.</li><li>RNOs integrate recurrent training into neural operator architectures to align training with inference dynamics.</li><li>Empirical results show that recurrently trained Multigrid Neural Operators outperform teacher-forced counterparts in long-term accuracy and stability.</li></ul>

Recurrent Neural Operators: Stable Long-Term PDE Prediction

Discover more